Proteoglycans in health and disease: the multiple roles of syndecan shedding

Authors


J. R. Couchman, Department of Biomedical Sciences, University of Copenhagen Biocenter, Ole Maaløes Vej 5, 2200 Copenhagen N, Denmark
Fax: +45 353 25669
Tel: +45 353 25670
E-mail: john.couchman@bric.ku.dk

Abstract

Proteolytic processes in the extracellular matrix are a major influence on cell adhesion, migration, survival, differentiation and proliferation. The syndecan cell-surface proteoglycans are important mediators of cell spreading on extracellular matrix and respond to growth factors and other biologically active polypeptides. The ectodomain of each syndecan is constitutively shed from many cultured cells, but is accelerated in response to wound healing and diverse pathophysiological events. Ectodomain shedding is an important regulatory mechanism, because it rapidly changes surface receptor dynamics and generates soluble ectodomains that can function as paracrine or autocrine effectors, or competitive inhibitors. It is known that the family of syndecans can be shed by a variety of matrix proteinase, including many metzincins. Shedding is particularly active in proliferating and invasive cells, such as cancer cells, where cell-surface components are continually released. Here, recent research into the shedding of syndecans and its physiological relevance are assessed.

Abbreviations
ADAM

a disintegrin and metalloproteinase

GAG

glycosaminoglycan

GlcA

glucuronic acid

GalNAc

N-acetylgalactosamine

GlcNAc

N-acetylglucosamine, HS, heparan sulfate

HSPG

heparan sulfate proteoglycan

MMP

matrix metalloproteinase

PKC

protein kinase C

PMA

phorbol myristate acetate

TIMP

tissue inhibitor of metalloproteinases

Introduction

Syndecans are type 1 transmembrane heparan sulfate proteoglycans (HSPGs) that have important roles during development, wound healing and tumour progression by controlling cell proliferation, differentiation, adhesion and migration. The heparan sulfate (HS) chains substituted on the extracellular domains interact with a wide range of ligands such as extracellular matrix glycoproteins, collagens, cytokines, chemokines, growth factors and enzymes, including metzincin proteinases. The ectodomain of each syndecan is constitutively shed in some cultured cells, but is accelerated in response to wound healing, and some pathophysiological events. Ectodomain shedding is an important regulatory mechanism, because it can rapidly generate soluble ectodomains that can function as paracrine or autocrine effectors or competitors. Mammals have four syndecan family members, syndecan-1 to -4 (Fig. 1), whereas invertebrates and primitive chordates possess only one syndecan, which is essential for neuronal development and axon guidance [1,2]. All cells express at least one member of the syndecan family [3], with the exception of erythrocytes. Syndecan-4 can be found in most tissues, but seems to be less abundant and is frequently coexpressed with other syndecans. Syndecan-1 is highly expressed in epithelia, syndecan-2 in endothelia and fibroblasts, whereas high expression of syndecan-3 can mostly be found in neuronal tissues and some musculoskeletal tissue. Here, our understanding of syndecan shedding and its function in wound healing and tumour progression is reviewed. Other reviews on syndecan structure and function have been recently published [4–6].

Figure 1.

 Schematic of the four vertebrate syndecans. Syndecans-1 and -3 core proteins are larger than those of syndecan-2 and -4, and can bear both heparan and chondroitin sulfate chains. The GAG chains are substituted on core protein serine residues and have a common stem tetrasaccharide of xylose (xyl), two galactose units (gal) and a glucuronic acid residue (GlcA). The repeating disaccharide of HS is N-acetylglucosamine and uronic acid, followed by several modifications in terms of sulfate and uronic acid epimerization to iduronic acid. The glucosamine can be N-, 6-O or (rarely) 3-O sulfated, whereas the iduronic acid can be 2-O sulfated. In most cases, there are regions of low sulfation, for example, adjacent to the core protein, with regions of intermediate or high sulfation. This yields a polysaccharide of immense variability and complexity. Chondroitin sulfate contains N-acetylgalactosamine, which may be 6-O or 4-O sulfated. The cytoplasmic domains have two highly conserved regions (C1 and C2) with an intervening syndecan-specific variable (V) region.

Structural organization of syndecans

The syndecan core proteins range from 20 to 40 kDa and have cytoplasmic domains that are highly conserved across species, but have diversity in their ectodomains. All comprise an ectodomain, a single transmembrane domain and a short cytoplasmic domain (Fig. 1). The cytoplasmic domain consists of membrane-proximal C1 and distal C2 conserved region flanking a variable region (V) that is unique to each syndecan, but highly conserved across species within each individual syndecan gene. The C2 region inter-acts with a number of PSD-95/Discs-large/Zonula occludens-domain-containing proteins such as syntenin, Gα-interacting protein (GAIP)-interacting C-terminus/synectin and calcium/calmodulin-associated serine kinase, since the C2 region contains a class II PSD-95/Discs-large/Zonula occludens protein-binding motif ΦXΦ, where Φ represent a hydrophobic residue and X any amino acid residue. Although information is sparse, current evidence suggests that the C1 region can interact with ezrin, at least for syndecan-2, which provides a link to the actin cytoskeleton [7]. The central V-region probably contains sites for syndecan-specific interaction partners, although this is only well understood for syndecan-4 [4,8]. A ternary signalling complex with phosphatidylinositol 4,5-bisphosphate and protein kinase Cα has been described [9], whereas others partners are the actin-associated protein α-actinin as well as syndesmos, about whose function rather little is known [10]. The transmembrane domain of all syndecans contains a GXXXG motif that promotes formation of SDS-resistant dimers [11,12]. The N-terminal ectodomain has glycosaminoglycan (GAG) chain substitution sites. These are predominantly HS covalently linked to serine residues in a serine–glycine motif surrounded by acidic residues. In addition to HS, syndecan-1 and -3 can be substituted with chondroitin or dermatan sulfate at sites closer to the transmembrane domain.

The synthesis of GAG chains in the Golgi apparatus is a highly complex process, but both HS and chondroitin sulfate chains are linked to serine residues on the core protein through a tetrasaccharide linker consisting of xylose–galactose–galactose–uronic acid residues, followed by the repeating disaccharide units. The repeating unit of HS and chondroitin sulfate backbones are glucuronic acid (GlcA)–N-acetylglucosamine (GlcNAc) or GlcA–N-acetylgalactosamine (GalNAc), respectively. These chains range from 50 to 200 disaccharides in length and undergo extensive modification in which some uronic acid residues are epimerized and a number of sulfation events occur (Fig. 1). In the case of HS, chain modifications are not uniform but localized along the chain. Subdomains of low sulfation are interspersed among regions that are highly sulfated, and small regions of intermediate sulfate lie at the boundaries of these subdomains [13,14]. How the synthesis of such complex polysaccharides is controlled remains unknown.

Syndecan shedding

Syndecans undergo regulated proteolytic cleavage, usually near the plasma membrane, in a process known as shedding. The release of syndecan extracellular domains may not only downregulate signal transduction, but also convert the membrane-bound receptors into soluble effectors/or antagonists. Soluble syndecan ectodomain can compete with intact syndecans for extracellular ligands in the pericellular environment [15] (Fig. 2). The remaining portion of the membrane-bound receptor loses its ability to bind ligands, and can be further processed by the presenilin/γ-secretase complex. Like many other type I transmembrane proteins [16], syndecan-3 has been shown to undergo restricted intramembrane proteolysis by the membrane presenilin/γ-secretase complex within the hydrophobic environment (mainly between Leu403 and Val404) of the phospholipid bilayer of the membrane [17]. In turn, there is decreased plasma membrane targeting of the transcriptional cofactor calcium/calmodulin-associated serine kinase. Signaling is not restricted to the syndecan proteoglycans but can be evoked by extracellular proteoglycans binding to cell-surface receptors. The leucine-rich proteoglycans are discussed in this context by Iozzo & Schaefer [18] in this minireview series.

Figure 2.

 Shedding of syndecans by metzincin proteinases. Several metzincin enzymes can cleave the syndecan core proteins, for example MMP9, the site(s) being membrane-proximal. Shedding is reported to be enhanced if the HS chains are first cleaved by heparanase. The shed syndecan may be deposited in the pericellular matrix, whereas the remnant core protein at the cell surface may be further processed by intramembrane cleavage by the presenilin/γ-secretase complex. There may also be signalling through MAP kinases.

Matrix metalloproteinases

Ectodomain shedding itself is a highly regulated process that requires the direct action of enzymes generally referred to as sheddases. All mammalian syndecan family members can be cleaved by extracellular protease [3]. The matrix metalloproteinases (MMPs) are known sheddases of syndecans, and are endopeptidases belonging to the family of metzincins (zinc endopeptidases) which contain three major multigene families: seralysins, astacins and a disintegrin and metallo-proteinase (ADAM)/adamlysins. Substrate specificity for MMPs is broad, therefore they function in many physiological processes and are key to normal matrix turnover, but also have essential roles in development and reproduction, and in pathological tissue remodelling during inflammatory disease, cancer invasion and metastasis. Normally, MMPs cleave substrates before a hydrophobic residue like Leu, Ile, Met, Phe or Tyr, whereas cleavage before a charged residue is rarely seen [19].

Twenty-three human MMPs have been identified which can be divided into eight distinct structural groups, five of which are secreted and three are membrane-bound (MT-MMPs) (Fig. 3). The general form of MMPs include an N-terminal signal sequence that directs them to the endoplasmic reticulum, a propeptide (Pro) containing a cysteine switch motif PRCGXPD (except for MMP23 which lacks the cysteine switch motif) that maintains them as inactive zymogens, and a catalytic domain with a zinc-binding site (Zn, HEXXHXXGXXH) and a conserved methionine (Met-turn) supporting the catalytic zinc. Interaction between cysteine–zinc maintains proMMPs in an inactive state by preventing a water molecule from binding to the zinc atom. All MMPs, with the exception of MMP-7, MMP-23 and MMP-26, also contain a hemopexin-like domain that is connected to the catalytic domain by a hinge region and mediates interactions with tissue inhibitors of metalloproteinases, cell-surface molecules and proteolytic substrates. The first and last of the four repeats in the hemopexin-like domains are linked by a disulfide bond (S–S) [19].

Figure 3.

 Schematic of mammalian matrix metalloproteinases. The domain structures of the various groups are shown, with a list of some members. S, signal peptide; Cat, catalytic domain; Pro, pro domain; TM, transmembrane domain; Cyt, cytoplasmic domain; Fu, furin cleavage site; Hpx, hemopexin domain; Fn, fibronectin type II repeats; V, vitronectin-like domain; CysR, cysteine array; Ig, immunoglobin-like domain, GPI, glycosylphosphatidylinositol linker.

Two gelatinase MMPs (MMP-2 and MMP-9) contain additional inserts that resemble collagen-binding type II repeats of fibronectin. MMP-11 and MMP-28 contain a basic amino acid motif [KX(R/K)R] recognized by furin-like serine proteinases between their propeptide and catalytic domains that results in their intracellular activation. This motif is also found in MMP-21 with the vitronectin-like insert (Vn), MMP-23 and the membrane-type MMPs (MT-MMPs) [19]. All soluble MMPs that do not harbour the basic motif at the end of propeptide are secreted as zymogens and activated extracellularly through proteolytic removal of propeptide. Active MMPs, plasmin, cathepsin G and neutrophil elastase have all been associated with this function. MT-MMPs can be subdivided into transmembrane (TM) forms and those that are glycosylphosphatidylinositol anchored. The TM-type MT-MMPs (MMP-14, MMP-15 and MMP-24) have a single-span transmembrane domain and a very short cytoplasmic domain. Alternately MMP-17 and MMP-25 are glycosylphosphatidylinositol-anchored MMPs. The type II membrane-linked MMP, MMP-23, has an N-terminal signal anchor targeting it to the cell membrane. Also, it is characterized by unique cysteine array and immunoglobulin-like domains.

In healthy adults, activity of MMPs is difficult to detect, except under conditions of tissue remodelling, for example, in wound healing and menstrual endometrium. Under physiological conditions, the activity of MMPs is regulated by transcription, activation of the precursor zymogen and by interactions with specific extracellular matrix components. In addition, endogenous tissue inhibitors of metalloproteinases provide a balance to prevent excessive degradation of extracellular matrix. This physiological balance may be disrupted in cancer. In many cancers, MMP expression is upregulated and correlates with poor prognosis [20,21]. Nevertheless, under some circumstances specific MMPs have a dual antitumour effect [22].

Tissue inhibitor of metalloproteinases

The catalytic activity of MMPs can be inhibited by the family of tissue inhibitor of metalloproteinases (TIMP), of which there are four members (TIMP1-4). TIMP-1, -2 and -4 are diffusible secreted proteins, whereas TIMP-3 is matrix associated because of its heparin-binding characteristics which promote its association with matrix proteoglycans [23,24]. TIMP-3 binds to sulfated glycosaminoglycans such as heparin, HS, chondroitin 4- and 6-sulfates, dermatan sulfate, and sulfated compounds such as suramin and pentosan, enabling interaction with GAG chains of syndecans [25]. Only TIMP-3 of the TIMP family has been shown to effectively block shedding of syndecan-1 and -4 in mouse mammary epithelial cells [26].

Each TIMP can inhibit most MMPs, except TIMP-1 that, in particular, fails to inhibit several of the membrane-type MMPs, MMP-14, -15, -16 and -24. The inhibitory effect of TIMP-3 is different from the others, as it also inhibits other metzincin subgroups, for example the ADAM/adamlysins, including ADAM-17 (TACE) [27], ADAM-10 [28] and ADAM-12 [29], and the ADAMs with thrombospondin motifs (ADAMTS) including the aggrecanases ADAMTS4 and ADAMTS5 [30]. Kinetic studies have shown that TIMP-3 is effective inhibitor of ADAM-17 (TACE) and aggrecanases [27,30]. All mammalian TIMPs consist of two distinct domains, N-terminal (∼ 125 amino acids) and C-terminal (∼ 65 amino acids), where the N-terminal domain usually is responsible for inhibition of proteinase activity. However, recently it has been shown that the isolated N-terminal domains of TIMP-1 and TIMP-3 are insufficient for ADAM10 inhibition, whereas full-length TIMP-1 and TIMP-3 are [31]. The C-terminal domain of TIMPs can stabilize proMMP by binding to its hemopexin domain, leaving the N-terminal fully capable of interacting with other MMPs. Most cell types secrete proMMP-9 in complex with TIMP-1, which complex can be found in the Golgi apparatus [32]. TIMPs -2, -3 or -4 can bind proMMP2, whereas TIMP-1 and -3 can interact with proMMP9.

TIMPs also facilitate activation of MMPs, by for example, functioning as an adaptor between MT1-MMP and Pro-MMP-2. MT1-MMP alone cannot bind proMMP2, but the N-terminal region of TIMP-2 binds the catalytic domain of MT1-MMP inhibiting its activity, whereas its C-terminal domain binds to the hemopexin-like domain of Pro-MMP-2 forming a ternary complex. The complexed MT1-MMP cannot cleave Pro-MMP-2, but requires a second MT1-MMP molecule (without TIMP-2). Thus cleavage and activation of proMMP-2 require both active and inactive MT1-MMP [33,34]. This process is facilitated by homodimerization of two MT1-MMP molecules through its hemopexin and transmembrane domains [35].

Syndecan sheddases

The glycosaminoglycan-bearing ectodomains of mammalian and Drosophila syndecans can be constitutively shed from the cell surface as part of the normal turnover [3,26,36–39]. This constitutive shedding involves metalloproteinases, but may be distinct from the metalloproteinase activity that mediates accelerated shedding in response to wound healing, for example [26].

Evidence indicates the involvement of several MMPs in syndecan cleavage in vitro and in vivo (Fig. 4). Matrilysin (MMP-7) cleaves syndecan-1 [40], gelatinases MMP-2 and MMP-9 can cleave syndecans-1, -2 and -4 [41,42], whereas the membrane-associated metalloproteinases MT1-MMP and MT3-MMP are known to cleave syndecan-1 [43]. However, current knowledge of precise cleavage-specific sites on syndecan core proteins is sparse. Human syndecan-4 is cleaved by the serine proteases, plasmin and thrombin, at Lys114–Arg115/Lys192–Val130 and Lys114–Arg115, respectively [44]. Despite high sequence homology between human and mouse syndecan-1, they have distinct MT1-MMP cleavage sites: human syndecan-1 is cleaved at Gly245–Leu246, whereas cleavage of mouse syndecan-1 occurs at Ala243–Ser244 [43,45].

Figure 4.

 Documented examples of metzincin proteinases that shed syndecans-1 and -4. Only in a few cases are the precise cleavage sites known. Most sites are believed to be membrane-proximal, although ADAMTS-1 and -4 may cleave syndecan-4 close to the N-terminus. CS, chondroitin sulphate; HS, heparan sulphate.

The ADAM family of disintegrin and metalloproteinase membrane-anchored proteinases [46] also participate in syndecan shedding. ADAM17 (TACE) has recently been reported to shed syndecan-1 and syndecan-4 [47]. The cysteine-rich domain of human ADAM12 was shown to associate with the ectodomain of syndecan-4 and is regulated by HS; however, direct ectodomain interactions with other members of the ADAM family are not known [48,49].

The ADAMTS family (disintegrin and metalloproteinase with thrombospondin motifs) [50] also associates with syndecans. It has been reported that the p53 form ADAMTS4 binds HS and chondroitin sulfate chains of syndecan-1 and aggrecan [51,52]. A recent study also reported that syndecan-4 may regulate activation of ADAMTS-5 via engagement of HS chains and regulation of MAPK-dependent synthesis of MMP3 during cartilage damage in osteoarthritis [53]. Therefore, lack of syndecan-4 may be chondroprotective in some models of osteoarthritis. Both ADAMTS-1 and ADAMTS-4 have been demonstrated to cleave syndecan-4 near the first GAG-attachment site, rather than close to the membrane. This was shown to decrease cell adhesion and promote cell migration [54].

Regulation of syndecan shedding

Syndecan shedding occurs through the direct action of sheddases, although a variety of extracellular stimuli including growth factors [55], chemokines [40,41,56], bacterial virulence factors [57,58], trypsin [36], insulin [59], heparanase [60] and cell stress [26] are known to induce syndecan shedding. It is not yet clear how extracellular stimuli influence sheddases to mediate syndecan cleavage, but different agonists appear to activate distinct intracellular signalling pathways to activate shedding. Chemical inhibitor studies suggest involvement of various signal transduction cascades, such as protein kinase C (PKC), protein tyrosine kinase, nuclear factor κB and mitogen-activated protein kinase pathways. For example, epidermal growth factor- and thrombin receptor-mediated shedding correlates with activation of the ERK/MAPK pathway, and does not appear to involve PKC activation. Inhibition of PKC activity prevents phorbol myristate acetate (PMA)- and cellular stress-induced shedding of syndecans, but does not affect thrombin or epidermal growth factor receptor-activated shedding [26,55].

Interestingly, some pathogens usurp the host cell shedding machinery to neutralize the host innate system to promote their own pathogenesis by elevation of syndecan shedding in response to bacterial virulence factors [61–63]. For example, Staphylococcus aureus, a common Gram-positive bacterium implicated in life-threatening diseases like endocarditis and osteomyelitis, enhances shedding of syndecan-1 through α-toxin and β-toxin [58]. Beta-toxin, but not α-toxin, also mediates shedding of syndecan-4. Alpha- and β-toxins do not directly trigger syndecan-1 shedding, but activate protein tyrosine kinase-dependent intracellular signalling pathways that stimulate syndecan-1 shedding [58]. Bacterial proteases can also enhance syndecan shedding by mimicking the direct shedding effect of syndecan sheddases [64]. For example, Streptococcus pneumoniae sheds syndecan-1 directly through ZmpC, a metalloproteinase virulence factor, where the size of the shed soluble ectodomain is smaller than that derived from α- or β-toxin mediated shedding [57]. Other pathogens may utilize HSPGs as attachment receptors to facilitate either their entry into the host cells or their survival in the host environment. For example, the capsid ORF2 protein of hepatitis E virus interacts mainly with 6-O-sulfate of syndecan-1 in Huh-7 liver cells for productive infection [65].

Intracellular regulatory mechanisms play important roles in agonist-induced shedding. Syndecans possess highly conserved transmembrane and cytoplasmic domains, the latter having three conserved tyrosine residues and a variable number of serine/threonine residues that can serve as phosphorylation sites [66]. Phosphorylation of tyrosine residues has been suggested to positively regulate syndecan-1 shedding [26,55,67]. The phosphatase inhibitor pervanadate and activation of intracellular kinases leads to tyrosine phosphorylation and shedding of syndecan-1 [68]. Hayashida et al. [69] confirmed the pervanadate effect on syndecan-1 shedding, but showed that S. aureusβ-toxin and PMA-mediated shedding was not accompanied by tyrosine phosphorylation. However, tyrosine to phenylalanine mutation reduced the syndecan shedding, suggesting mechanisms other than phosphorylation, such as binding to other cytoplasmic components is critical in agonist-mediated shedding. For example, syndecan-1 cytoplasmic domain interacts with the inactive, GDP-bound form of Rab5, a small GTPase that regulates intracellular trafficking and triggers its conversion to an active GTP-bound state in response to shedding promoters. A dominant negative form of Rab5, unable to switch between active and inactive states, significantly inhibited syndecan-1 shedding, suggesting that trafficking is a key regulator of syndecan-1 shedding [69].

Wound healing

Wound healing is a regulated process that can be divided into three sequential, yet overlapping, phases; inflammation, proliferation and remodelling [70]. Syndecan-4 is upregulated in a range of inflammatory conditions like ischaemic myocardial injury [71], and dermal wound repair [72]. For example, atherosclerosis is a chronic inflammatory disease marked by aberrations in cell migration, proliferation and low-density lipoprotein internalization [73]. Oxidized linoleic acid, the major oxidized fatty acid in low-density lipoprotein, upregulates expression of syndecan-4, and as a consequence, accelerated shedding of syndecans-4 involving the MEK pathway [74]. Increased levels of syndecan-1 ectodomain are present in dermal wound fluid, and in serum from patients with acute graft-versus-host disease [75].

A key inflammatory response is chemokine-mediated recruitment of leukocytes into sites of inflammation [76]. Many chemokines bind HS chains of syndecans and evoke MMP-mediated shedding of syndecans with potential loss from the site of injury [40,41,56]. MMP-7 is upregulated in injured mucosal epithelium of the lung, and promotes inflammation by shedding a syndecan-1/KC (CXCL8) complex that directs neutrophil influx to the sites of injury [40]. Soluble syndecan-1 may maintain the proteolytic balance of acute wound fluids, because it can bind the inflammation-related neutrophil proteases cathepsin G and elastase, consequently decreasing their affinity for their physiological targets [37].

The function of syndecan-1 shedding in wound healing is not restricted to inflammation, but serves also to promote re-epithelialization; however, this is not fully clarified. Proliferating keratinocytes at the wound edge and endothelial cells in the wound bed transiently express syndecan-1 [77], whereas keratinocytes migrating into the wound lose their cell-surface syndecan-1 expression [37]. Syndecan-1 and syndecan-4 are shed and may accumulate in dermal wound fluids [55]. Using a noncancerous simple epithelium cell line (BEAS-2b) and organotypic cultures derived from primary epithelial cells, it has been demonstrated that syndecan-1 is shed primarily by MMP-7 from epithelial cells after injury [78], which enhances cell migration and facilitates wound closure. Therefore, syndecan-1 shedding appears to be an important response in wound healing. MMP-7 null mice demonstrate a severely diminished re-epithelialization in response to lung injury. Suppression of syndecan-1 expression in simple epithelial cells induces a promigratory phenotype [79,80], consistent with decreased syndecan levels in injured stratified epithelium (cornea and skin) during repair [81,82]. Furthermore, knockdown of syndecan-1 expression resulted in slowed cell migration in an A549 (a carcinoma-derived alveolar type II) cell line [83]. Interestingly, soluble syndecan-1 ectodomain inhibited wound repair in mice overexpressing syndecan-1, by exhibiting delay in wound closure, re-epithelialization, granulation tissue formation and remodelling [84]. Overall, the studies reveal that MMP-7 cleavage of syndecan-1 is essential for effective re-epithelialization; however, a balance is critical because soluble syndecan-1 overexpression or complete absence of syndecan-1 in the knockout lead to impairment. The function of syndecan-1 may be tissue specific, because syndecan-1 null primary dermal fibroblasts migrated faster than wild-type cells [85]. E-cadherin, a known mediator of cell–cell contact, is also shed in vivo from injured lung epithelium by MMP-7 [86], and has been shown to be coordinately regulated with syndecan-1 [79]. It is not known if shedding of E-cadherin and syndecan-1 happen contiguously, but could synergistically promote a migratory epithelial phenotype.

It is well known that syndecans are functionally coupled to integrins [4], which represent the major group of cell-surface receptors for extracellular matrix macromolecules. There are 24 heterodimeric integrins in mammals, each composed of an α and a β subunit derived from combinations of 8 β and 18 α subunits. Interaction between syndecan and integrins may be direct [87] or indirect through an intermediate ‘receptor’ [88]. This adhesion mechanism can be HS independent, because the cell adhesion properties of syndecans are not only limited to the HS chains, but can also be mediated through the ectodomain core protein. The evolutionarily conserved NXIP motif of syndecan-4 has been shown to promote β1-integrin-dependent cell adhesion [89]. Syndecan-1 ectodomain regulates αvβ3 and αvβ5 integrin-mediated attachment and spreading in human mammary carcinoma cells and B82L fibroblasts, respectively. The activity has been mapped to residues 88–252 within the syndecan-1 ectodomain [90,91]. This association can be blocked by synstatin, a peptide inhibitor corresponding to the active site of the syndecan-1 core protein, and which can suppress angiogenesis in vitro and in vivo, perhaps signifying syndecan-1 as a critical mediator of tumour progression [87].

Another motif, the AVAAV (amino acids 222-226), only present within the syndecan-1 ectodomain, has been suggested to be an invasion regulatory domain, because mutation within this region abolishes syndecan-1-mediated inhibition of cell invasion [92]. However, the mechanism remains unknown.

Integrins and syndecans together may influence the outcome of cell adhesion and migration because their different activation states and clustering on the cell surface result in varying degrees of mechanical force exerted on the extracellular matrix [5]. Syndecan-1 shedding by MMP-7 from repairing simple epithelial (BEAS-2b) cells after injury [77] enhances cell migration and facilitates wound closure by causing the α2β1 integrin to assume a less-active conformation, compatible with migration. It has previously been shown that syndecan-1 facilitates integrin α2β1-mediated adhesion to collagen [93].

Tumour progression

In addition to genetic and epigenetic changes, tumour progression links a series of steps involving adhesion, motility and growth, resulting in metastatic spread, a major cause of death among cancer patients. These steps are influenced by the activity of tumour-derived MMPs. MMPs facilitate metastasis by degrading extracellular matrix components, such as collagens, laminins and proteoglycans, and they modulate cell adhesion, enabling turnover of matrix contacts or adhesions. Novel roles for proteoglycans in malignancy are also discussed elsewhere in this volume [94].

As part of the regulation of MMPs, rate-limiting effects, such as zymogen activation and the availability of TIMPs are important. Another control element may be contributed by HS chains of proteoglycans, which interact with many extracellular protease, with examples from all four classes of proteases (aspartyl-, seryl-, cysteyl-protease and metalloproteases). Heparan sulfate also interacts with protease inhibitors, for example TIMP-3 [95] and antithrombin III (ATIII). These interactions may control extracellular matrix degradation, by either modifying enzymatic activity through activation or inhibition, or providing a reservoir of latent enzyme that is positioned for directed proteolytic attack on extracellular matrix proteins. For example, highly sulfated HS has been shown to inhibit the proteolytic degradation of aggrecan, in part through direct inhibition of aggrecanase activity [96]. Furthermore, HS chains of syndecans bind tumour-associated MMPs, MMP-2, -7, -9 and -13 [97], in which MMP-2 catalytic activity is inhibited by its interaction with HS chains of syndecan-2 [98], whereas MMP-1, -7 and -13 catalytic activity increases in the presence of heparin [97]. MMP-7 has been shown to promote syndecan-1 shedding upon growth factor activation (FGF-2), achieving its own release although still being attached to HS chains [97]. Other attributes of HS chains include the ability of TIMP-3 to interact with cell-surface HS. This may lead to inhibition or internalization of cell-surface MMPs or ADAMs, because it has been discovered that TIMP-3 is internalized in HEK293 and HTB94 chondrosarcoma cells [99], a process that is mediated by cell-surface glycosaminoglycans [99,100].

Overall, HS chains of syndecans may support invasion of tumour cells by protecting and anchoring matrix-degrading proteases, while also harbouring signalling molecules that promote growth and directional migration. However, the MMP-13 C-terminal domain has been shown using yeast two-hybrid analysis to associate with syndecan-4 without HS chains, suggesting alternative MMP interaction sites than GAG chains [101].

The role of syndecans in tumour progression may vary with tumour stage and type, because syndecan-1 is reported to be downregulated in several types of breast cancer [102], but upregulated in several tumours, such as pancreatic cancer. Soluble syndecan-1 ectodomain can be found in the serum of lung cancer patients [103] and Hodgkin’s lymphoma patients [104], in the extracellular matrix of myeloma biopsies, as well in the serum of myeloma patients [105,106], to a much greater degree than in healthy individuals [107].

A recent study has distinguished the roles between membrane-bound and shed form of syndecan-1 in breast cancer epithelial cells (MCF-7) in vitro. The membrane-bound form of syndecan-1 increased proliferation and inhibited invasiveness, whereas the soluble form had the opposite effect, by promoting invasiveness and inhibiting proliferation [108].

Perhaps the best evidence for the importance of shedding in cancer is shown for syndecan-1 in myeloma. Multiple myeloma is a malignant proliferation of the bone marrow plasma cells increasing angiogenesis and development of osteolytic bone disease. Soluble syndecan-1 promotes the growth of myeloma tumours in vivo [109]. High levels of shed syndecan-1 in the sera of myeloma patients are a marker of poor prognosis [105,107,110].

Heparanase seems to play a distinct role in shedding syndecans in myeloma. Mammalian heparanase (endo-β-d-glucuronidase) is known to modulate syndecans by cleaving the less-sulfated regions along the HS chain releasing fragments of 10–20 sugar residues [111] (Fig. 2). It may function in tumour progression by promoting tumour growth, angiogenesis and metastasis [112] by both enzymatic and nonenzymatic mechanisms. A recently described nonenzymatic mechanism of heparanase is its ability to facilitate cell adhesion and spreading by clustering of syndecan-1 and syndecan-4 through interaction with their HS chains [113]. Knockdown of heparanase in myeloma cell lines decreases soluble syndecan-1 [114]. In support, active heparanase was shown to accelerate myeloma cell growth and promote bone metastasis by increasing the number and size of blood vessels within the tumour [115,116]. Heparanase function in tumour progression is discussed by Barash et al. [117] in this minireview series.

Elevated active heparanase has been demonstrated to enhance syndecan-1 shedding through ERK signalling, which in turn upregulates expression of two proteases, MMP-9 and urokinase-type plasminogen activator [118]. Recently, it has been shown that heparanase-enhanced shedding of syndecan-1 by myeloma cells promoted endothelial invasion and angiogenesis [118]. Heparanase also increased urokinase-type plasminogen activator receptor expression levels [119], and can even initiate syndecan-1 expression in the ARH-77 (human plasma cell leukemia) cell line that is normally negative for syndecan-1 [60]. The expression of urokinase-type plasminogen activator and its receptor may also be a predictor of poor prognosis, just as with shed syndecan-1 and heparanase [120].

The gelatinase MMP-9 sheds syndecan-1 directly [41], and has been suggested as a useful prognostic index of bone disease [121]. In addition, myeloma cell invasiveness can be promoted by MMP-9 in vitro [122], consistent with data suggesting that MMP-9 inhibition has antimyeloma effects [123]. Urokinase, by contrast, has a more indirect effect on syndecan shedding. Its activity in generating plasmin from plasminogen has been suggested to be a major activator of MMPs in vivo, where it can process proMMP into active MMP. In turn, these shed syndecans directly and/or activate other MMP sheddases. For example, plasmin directly activates proMMP-1, proMMP-3, proMMP-9, proMMP-10 and proMMP-13 in vitro [124].

Conclusions and perspectives

Syndecan shedding is subject to highly complex regulation. In tissue culture, there may be constitutive shedding, and in vivo enhanced shedding in cases of injury and disease. Because syndecans are important co-receptors for adhesion and growth factor receptors, their loss from the cell surface may have multiple effects. There is certainly a need for a deeper understanding of these processes, because they may relate to diagnosis, prognosis or even treatment options for some diseases. Better reagents for detecting syndecan cleavage will be a valuable aid in these analyses, both in vitro and in vivo. This may be difficult, not least because so many different proteases can cleave the syndecan core proteins. There is much to learn about when and where these events take place.

Ancillary

Advertisement